面向时序可预测传感器处理系统的GPU驻留执行
GPU-resident Execution for Timing-predictable Sensor Processing Systems
AI总结:
本文提出GPU驻留执行模型,在NVIDIA Holoscan C++框架实现,可降低实时传感器处理系统的抖动,在1080p 60Hz摄像头应用中使光子到显示抖动降7.3倍、软件处理抖动降170倍。
摘要:
实时嵌入式系统日益将高带宽传感器流的GPU加速处理纳入闭环系统,但CPU(主机)仍为每次迭代的主要协调者:主机观测传感器输入、提交CUDA任务、等待完成并执行驱动。即使采用GPU内存RDMA或CPU核心隔离等常见主机端缓解措施,此主机介导路径仍使闭环端到端延迟暴露于I/O中断、主机调度器决策及CUDA API开销中。本文提出GPU驻留执行,一种用于实时传感器处理流水线的新型执行模型,其将I/O与GPU执行置于长期设备驻留循环中:传感器输入抵达GPU内存后触发GPU侧处理,最终到达GPU可见输出端点,无需任何主机唤醒。我们在NVIDIA Holoscan C++框架中实现该模型,并提供构建GPU驻留应用的参考组件。我们在NVIDIA GPU与SoC上针对推理、主机竞争、多循环及流摄像头工作负载开展评估:在工业级1080p、60Hz实时摄像头到显示应用中,与CPU驱动的GPU内存RDMA基线相比,GPU驻留执行将光子到显示抖动从4.83ms降至0.66ms(降低7.3倍),并将软件处理抖动从1.70ms降至0.01ms(降低170倍)。
英文摘要:
Real-time embedded systems increasingly place GPU-accelerated processing of high-bandwidth sensor streams inside closed-loop systems, but the CPU (the host) remains the primary orchestrator of each iteration. The host observes sensor inputs, submits CUDA work, waits for completion, and commits actuation. This host-mediated path exposes the closed-loop end-to-end latency to I/O interrupts, host scheduler decisions, and CUDA API overheads even after common host-side mitigations such as GPU-memory RDMA or CPU core isolation. This paper presents GPU-resident Execution, a new execution model for real-time sensor processing pipelines that keeps I/O and GPU execution in a long-lived device-resident loop. Sensor inputs arrive in GPU memory, trigger GPU-side processing, and reach GPU-visible output endpoints without any host wakeups. We implement the model in the NVIDIA Holoscan C++ framework and provide reference components for building GPU-resident applications. We evaluate our system on NVIDIA GPUs and SoCs with inference, host-contention, multi-loop, and streaming camera workloads. On an industry-grade 1080p, 60 Hz real-time camera-to-display application, GPU-resident Execution reduces photon-to-display jitter from 4.83 ms to 0.66 ms (7.3x) relative to a CPU-driven GPU-memory RDMA baseline and reduces software-processing jitter from 1.70 ms to 0.01 ms (170x).
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。